Re: Oracle row limit support (again)

Hi Tomas, This is wonderful. This is what I like and admire - simple brilliant solutions. No parsing of the column names! Regards, John PS: My algorithm for ADOdb's SelectLimit( ) will be something like this in the future... if $offset < 100 then ### hmm, few rows, so no need to execute 2 SQL select's select * from ($sql) else ### use Tomas algorithm (provided he doesn't patent it ### and asks for royalties !) endif Tomas V.V.Cox <cox@idecnet.com> wrote in message news:3C06EA19.3472D091@idecnet.com... > Well, finally I thought in a better solution for the problem. Instead of > breaking my head with the SQL parser and trying to guess the resultant > columns for building the special query needed by the limit stuff, I used > a nice trick: launch the real query but with a special clausule (WHERE > NULL = NULL) that won't return anything execept the field names I need. > > I haven't more time to test that with many queries, but looks good and I > think it should work with any query. > > The results are impressive according to my benchmarks: > > Bench info: > P233MMX - Linux - Oracle 8.0.5 - PHP cgi 4.1 > Table with 10.000 rows > Always fetching 20 rows from the start position. > No average calculation, taken the first results returned > > Leyend: > OUT (Throwing out the data until Oracle reachs the correct position) > PDB (Using the PEAR DB Oracle row limit) > > From = 0 > PDB: 0.019351 > OUT: 0.011932 > > From = 1000 > PDB: 0.025988 > OUT: 1.191379 > > From = 5000 > PDB: 0.016395 > OUT: 5.539404 > > From = 9500 > PDB: 0.023285 > OUT: 11.037733 > > > Tomas V.V.Cox > > PD.- John, perhaps you want to add this to your benchmarks? ;-)